Testing for Trend in Benefit-Risk Analysis with Prioritized Multiple Outcomes

Benefit-risk analysis using prioritized multiple outcomes has been proposed for use in randomized controlled trials with two treatment groups. This research extends the two-group comparison to testing for a trend over multiple treatment groups, for example, multiple doses, in terms of a composite outcome that is derived from pairwise comparisons of subjects between groups or by ranking subjects of pooled groups, according to their prioritized outcomes. Tukey’s trend test, Wilcoxon rank-sum test and Jonkheere-Terpstra (JT) test and their permutation-based trend tests are investigated for detection of an increasing trend over doses with respect to the composite benefit-risk endpoint. Simulation studies show that the permutation-based Tukey’s and JT tests outperform the others in terms of Type I error control and power under various simulation settings. For illustrative purpose, the six approaches are applied to a migraine example data to determine whether an increasing trend exists among four dose groups in terms of a composite benefit-risk endpoint that are measured by four prioritized outcomes.

针对两治疗组随机对照试验（randomized controlled trial），已有研究提出采用优先排序多结局的获益-风险分析（benefit-risk analysis）方法。本研究将该方法从两群组比较场景拓展至多治疗组（例如多剂量组）的趋势检验场景，其综合结局通过组间受试者两两比较，或对合并群组内的受试者基于其优先排序结局进行排名的方式构建。针对基于综合获益-风险结局的剂量递增趋势检测任务，本研究对Tukey趋势检验、Wilcoxon秩和检验、Jonckheere-Terpstra（JT）检验及其基于置换的同类趋势检验方法展开了性能评估。仿真实验结果表明，在各类仿真场景下，基于置换的Tukey检验与JT检验在Ⅰ类错误控制与检验效能方面均优于其余方法。为展示方法应用流程，本研究将六种检验方法应用于偏头痛示例数据集，基于四项优先排序结局构建综合获益-风险结局，以此判断四个剂量组间是否存在递增趋势。